65-nm CMOS, W-band Receivers for Imaging Applications

Size: px

Start display at page:

Download "65-nm CMOS, W-band Receivers for Imaging Applications"

Clifford Oliver
5 years ago
Views:

1 65-nm CMOS, W-band Receivers for Imaging Applications Keith Tang Mehdi Khanpour Patrice Garcia* Christophe Garnier* Sorin Voinigescu University of Toronto, *STMicroelectronics University of Toronto 27 1

2 Table of Content Motivation Circuit Schematics Fabrication Measurement Results Conclusion University of Toronto 27 2

3 Motivation Investigation of W-band receivers in 65-nm GP CMOS CMOS might provide alternatives to III-V and SiGe technology in imaging arrays: Broadband (multi-ghz) Low noise Low power Small area Comparison of two LNA feedback topologies Series-series feedback with inductor Shunt-series feedback with transformer University of Toronto 27 3

4 Receiver Block Diagram University of Toronto 27 4

5 LNA Schematic Inductive (series-series) feedback LNA Input matched by L G and L S R{ Z } = 2πf L + R + Noise impedance matched by transistor sizing and biasing IN T S G R S University of Toronto 27 5

6 LNA Schematic (2) Transformer (shunt-series) feedback LNA LP Input matched by L P, L S and M R { Z IN}, M = k LPL gm M Noise impedance matched by transistor sizing and biasing S University of Toronto 27 6

7 LNA Simulation [db] ind-feedback -5-5 [db] xfmr-feedback S 21 S 11 Γ opt S 21 S 11 Γ opt NF 5 NF MIN NF 5 NF MIN FREQUENCY [GHz] FREQUENCY [GHz] S 11, Γ opt < -1 db from 74-1 GHz for both designs University of Toronto 27 7

8 Mixer Schematic Gilbert cell mixer with inductive broad-banding University of Toronto 27 8

9 Fabrication 65-nm GP/LP digital CMOS process 7 metal layers GP n-mosfets (8 6nm 1μm) with gate contacted on one side: f T /f MAX =17 GHz/2 GHz at V DS =.7 V GP MOSFETs 3% faster than LP MOSFETs and require lower V GS and V DS lower power Gate leakage does not affect mm-wave performance University of Toronto 27 9

10 LNA breakouts Die Photos XFMR IND IND-feedback XFMR-feedback 49 um x 3um (pad) 12 um x 17 um (core) University of Toronto 27 1

11 Mixer breakout Die Photo LO IN IF P IF N RF IN 47 um x 56 um (pad) 19um x 16 um (core) University of Toronto 27 11

12 Receiver Die Photos IND-feedback Receiver XFMR-feedback Receiver 46 um x 5 um (pad) 16 um x 37 um (core) University of Toronto 27 12

13 Meas. LNA 1 st Spin S 21, S 11 (db) ind-feedback S 21, S 11 (db) xfmr-feedback S 21 (sim.) -2 S 21 (meas.) -25 S 11 (sim.) -25 S 11 (meas.) FREQUENCY (GHz) Requires 2.2 V for 8 9 db gain 4 5 db below simulation -2 S 21 (sim.) -2 S 21 (meas.) -25 S 11 (sim.) -25 S 11 (meas.) FREQUENCY (GHz) University of Toronto 27 13

14 Measurements for 2 nd Spin with Modified Layout Series resistance in ground metallization of LNA was found in the first spin. A second spin of the design was fabricated with: Wider metal lines in ground mesh at top level Increased number of vias (even between M5 and M6) LNA inductance values adjusted to 8 GHz University of Toronto 27 14

15 22 2 Meas. LNA 2 nd Spin =1.8V (1 st spin) =2.2V (1 st spin) 22 2 =1.8V (1 st spin) =2.2V (1 st spin) =1.2V (2 nd spin) =1.5V (2 nd spin) =1.8V (2 nd spin) =1.2V (2 nd spin) =1.5V (2 nd spin) =1.8V (2 nd spin) GAIN [db] 12 1 GAIN [db] ind-feedback FREQUENCY [GHz] FREQUENCY [GHz] Measured 1.5 V = 13 db 2 xfmr-feedback University of Toronto 27 15

16 2 nd Spin LNA meas. vs sims S 21, S 11 (db) ind-feedback S 21, S 11 (db) xfmr-feedback S 21 (sim.) -3 S 11 (sim.) -3 S 21 (meas.) S 11 (meas.) FREQUENCY (GHz) S 21 (sim.) -3 S 11 (sim.) -3 S 21 (meas.) S 11 (meas.) FREQUENCY (GHz) Meas. = 1.5 V is 1 2 db below sims. S 11 < -2 db from 8 9 GHz (xfmr-feedback) University of Toronto 27 16

17 [db] Meas. Transformer S-params S 11 S 22 S MAG FREQUENCY [GHz] MAG (loss) < -2 db between GHz University of Toronto 27 17

18 Meas. Mixer Conversion Gain 9 CONVERSION GAIN [db] =1.5V =1.8V 1 =1.8V (sim.) RF FREQUENCY [GHz] IF = 1GHz 1 2 db below simulation University of Toronto 27 18

19 Meas. Mixer NF DSB IF = 1GHz NOISE FIGURE [db] RF FREQUENCY [GHz] NF MIXER, =1.5V NF MIXER, =1.8V Includes ~2 db transformer loss Lowest NF DSB mixer at 8 9 GHz in silicon University of Toronto 27 19

20 Meas. Rx Gain, NF DSB vs IF CONVERSION GAIN [db] =1.2V =1.5V =1.8V IF FREQUENCY [GHz] NF DSB [db] =1.2V =1.5V =1.8V IF FREQUENCY [GHz] XFMR-feedback RCVR NF DSB ~7 8 db, 89 GHz University of Toronto 27 2

21 Meas. Rx Gain, NF DSB vs RF 15 IF = 1GHz 2 1 GAIN, NF [db] S 11 [db] GAIN RCVR NF DSB RCVR -3 S RF FREQUENCY [GHz] 3dB-bandwidth: GHz University of Toronto 27 21

22 Receiver P 1dB GAIN [db] LO = 75GHz RF = 8GHz P 1dB = -16.2dBm P IN [dbm] 75 GHz due to equipment limitation 5-5 P OUT [dbm] University of Toronto 27 22

23 Estimated LNA NF [db] GAIN, =1.5V 4 GAIN, =1.8V 2 NF 5, =1.5V 2 NF 5, =1.8V FREQUENCY [GHz] XFMR-feedback LNA G F LNA LNA = G = F RCVR RCVR G F G MIXER MIXER LNA 1 LNA gain peaks at frequency higher than measured (output pad capacitance removed) LNA NF 5 ~6 7 db University of Toronto 27 23

24 Summary of Results 1 st Spin [V] P diss LNA IF Buffer Receiver Gain P diss P diss Gain NF S 11 [mw] [db] [mw] [mw] [db] [db] [db] < ( GHz) 2 nd Spin < ( GHz) Dramatic increase in performance just with better top-level ground mesh and vias University of Toronto ~ ½ of P diss used in IF buffer to drive 5Ω off-chip

25 Conclusion GHz receiver with 8 db NF and 13 db gain demonstrated in 65 nm GP CMOS technology. Inductive-feedback and transformer-feedback LNA topologies presented: Similar performance achieved by different matching procedures Layout style significantly affects circuit performance. Post-layout simulation at top-level, with ground mesh must be carried out. University of Toronto 27 25

26 Acknowledgement Katya Laskin for measurements on the second-spin Alex Tomkins for inductor and transformer measurements Jaro Pristupa and CMC for CAD tools Bernard Sautreuil of STM for facilitating the technology access CITO for funding ECTI, NSERC, CFI and OIF for equipment University of Toronto 27 26

27 2 nd spin LNA Meas. Gain S 11, S 21 [db] S 11, S 21 [db] =1.2V -2-2 =1.2V =1.5V =1.8V =1.5V =1.8V FREQUENCY [GHz] FREQUENCY [GHz] IND-feedback LNA XFMR-feedback LNA S 11 matched at 93 GHz for inductive-feedback LNA (increase L G ) University of Toronto 27 27

28 CONVERSION GAIN [db] Receiver vs LO Power LO = 85GHz RF = 84.5GHz ind-feedback LO POWER [dbm] =1.8V =2.V =2.2V CONVERSION GAIN [db] LO = 85GHz RF = 84.5GHz xfmr-feedback =1.8V =2.V =2.2V =1.2V (2 nd spin) =1.5V (2 nd spin) =1.8V (2 nd spin) LO POWER [dbm] Requires 2 3 dbm (1 st spin) and > 5 dbm (2 nd spin) LO power University of Toronto 27 28

29 5 Receiver P 1dB LO=77GHz, RF=75GHz LO=77GHz, RF=75GHz P OUT [dbm] -1 6 GAIN [db] P OUT [dbm] -1 6 GAIN [db] P IN [dbm] P IN [dbm] IND-feedback Receiver XFMR-feedback Receiver RF at 75 GHz due to equipment limitation University of Toronto 27 29

95GHz Receiver with Fundamental Frequency VCO and Static Frequency Divider in 65nm Digital CMOS

95GHz Receiver with Fundamental Frequency VCO and Static Frequency Divider in 65nm Digital CMOS Ekaterina Laskin, Mehdi Khanpour, Ricardo Aroca, Keith W. Tang, Patrice Garcia 1, Sorin P. Voinigescu University